Google: attackers now use autonomous multi-agent AI frameworks in live intrusions
Google Threat Intelligence Group (GTIG) and Mandiant published a new edition of their AI Threat Tracker on 16 November 2026. Their conclusion: adversarial use of AI has reached an inflection point, as threat actors move from manual prompting to autonomous multi-agent frameworks that chain intrusion steps at machine speed.
The report names open source offensive frameworks such as ARTEX, CAIRN, CyberStrikeAI and PentAGI. GTIG observed CAIRN in intrusions against US aerospace targets, and tracked an actor running a customised PentAGI platform for reconnaissance and scanning against organisations in Asia between late August and early October 2026. Other trends include automated reverse engineering of patches to weaponise n-day vulnerabilities, « LLMJacking » (theft and resale of access to AI models), routine use of LLMs in influence operations, including by commercial influence-as-a-service vendors, and evidence that a PRC-nexus group was building its own automated attack framework for likely espionage use.
For economic security, the main shift is speed: known weaknesses such as stolen credentials and unpatched systems can now be exploited faster, which shortens the time defenders have to react. As a vendor publication, the report draws on Google’s own telemetry and also promotes its defensive products.
Source: Google Cloud Blog (GTIG)
Image: VGrigas (WMF) / Wikimedia Commons (CC BY-SA 3.0)
